Greetings - I have been catching up on the announcements made during Cloudflare’s Platform Week that occurred May 9-14. There were a lot of new developments, more than in prior innovation weeks, illustrating Cloudflare’s accelerating pace of innovation. I thought I would take a minute and share the announcements that I thought were particularly interesting, which could impact the adoption of Cloudflare’s development platform and create future revenue streams. For this post, I will focus on the data products. There was a good thread on the board earlier about D1: https://discussion.fool.com/cloudflare-announces-d1-database-351…. In addition to D1, I thought Cloudflare announced a few other data product releases with potential.
At a high level, what I find most interesting is Cloudflare’s rapid expansion into data storage, processing and distribution. With their second data storage offering (Durable Objects) launched just two years ago, Cloudflare has quickly added object storage, a basic relational database, a time series database and event messaging to the mix. These all represent tools that a developer would expect to build a functional Internet application. By making all of these solutions available in parallel from all data center locations, these capabilities add application use cases that could be addressed from the edge (generally those requiring state management). Granted, not every application gains noticeable benefits from running at the edge, but those with latency sensitivity would. Some of these new products are in a basic form, but Cloudflare has demonstrated the ability to launch a beta product, get customer feedback and then expand capabilities rapidly (much like some other board favorites like Datadog).
I think this trend of product expansion in data services will continue for Cloudflare. Each of these new products will be monetized, providing an incremental revenue source going into 2023. For R2 alone, I did some back of the envelop math on what “hundreds of petabytes” of storage (quote from Q4 earnings call) would represent with published pricing and came up with about $30M of annual revenue. Two years ago, I wouldn’t have modeled any revenue for Cloudflare from data storage and processing. Over the next 1-2 years, data products could contribute a meaningful share.
With that said, these are the product announcements related to data storage that I found most interesting:
-
R2. This is Cloudflare’s object store, similar to Amazon’s S3. With this announcement, the product moved from closed beta to open beta. GA is expected in Q4. They published pricing, which is about 10-30% cheaper than S3, and with no egress fees. They also specified thresholds for free usage with the open beta, but implied that usage over these levels would be billed. Cloudflare does have work remaining to fully scale this solution - load is being capped at 1k reads and 100 writes per second per bucket. They intend to raise these thresholds as they get more operational experience with the platform. Initial feedback from enterprise customers is that they see R2 as an opportunity to cheaply store bulk data and distribute it across clouds and between partners.
-
D1. This was probably the highlight of the Week. D1 is a SQL compatible database, built on top of the open source embedded database SQLite. This represents a smart choice, as Cloudflare is able to quickly spin up a SQL interface and leverage a highly performant, well-understood solution. D1 will perform like any relational database. It is ACID compliant, contains all the standard SQL commands, is well-tested and fast. D1 is built on top of Durable Objects as the storage engine. This provides the consistency guarantee, as well as a caching layer. Developers can reference D1 from any Worker instance, without having to designate a location for the data (an advantage over hyperscaler edge functions like Lambda@Edge). Cloudflare manages the routing, caching optimization and replication behind the scenes.
-
Pub/Sub. While D1 grabbed the headlines, I think Pub/Sub is a high potential capability. This adds a programmable messaging service to the platform. Publish-subscribe is a common pattern that allows multiple entities to generate messages and distribute them to many consumers. Messages can represent events or sensor data. The protocol is based on MQTT, which is an industry standard generally associated with IoT data distribution. Common use cases include distribution of sensor readings, telemetry, financial transactions and mobile notifications. Because the Pub/Sub service will be accessible from any Worker in all data centers, it will provide an efficient system for collection and distribution of data from smart devices that are deployed across the globe. Hyperscaler solutions generally assume that data collection and distribution are centralized within a specific region.
-
Workers Analytics Engine. On the surface, this provides a simple data store and query engine for metrics data (like observability). Under the covers, though, it is implemented as a time series database. Data is written as “data points”, which consist of labels and metrics (data values). Once stored, the data can be retrieved using a SQL API, generally filtered by labels. This provides a convenient destination for IoT sensor data when combined with Pub/Sub, providing more support for distributed IoT data processing.
With these products, the Cloudflare development platform now has a key-value store, an object store, a relational database, a times series database and a publish-subscribe system. I think these combine to provide a fairly complete set of capabilities for developers. These may entice more enterprises to move parts of application workloads (peel off latency sensitive services and UI functions) to run on Cloudflare Workers. As part of the Q1 report and Investor Day, we got highlights of a couple of new wins - Meta signed a $3M / 5 year deal to host an authentication service on Workers, Atlassian is running a collaboration feature for $145/year (expected to grow from here). These are material contracts and I expect more announcements like these in 2023. During Investor Day, they also announced that 450k developers have used Workers so far. They expect to more than double this count to 1M by the end of this year. That would represent a pretty significant ramp in adoption.
Of course, the actual impact on revenue growth for the remainder of 2022 remains to be seen. Additionally, these new products are requiring a ramp in CapEx spend, which is impacting FCF. Personally, I am comfortable with that investment, as I think their hardware utilization will be very high and like the advantages Cloudflare will gain by owning their infrastructure. However, I also appreciate that this makes the financials harder to rationalize side-by-side with some other rapid growers that are leveraging hyperscaler infrastructure to minimize capital spend.
Cheers,
Peter Offringa, Software Stack Investing, @StackInvesting (NET is top 3 position)